الفهرس 
CHAPTER (1): INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 150 |  |  |
| | | from | 150 | | |
Abstract
Millions of cars and trucks travel thousands of kilometers of paved roads
every day in Egypt to transport people and goods. Pavement deterioration occurs
over time as a result of traffic volume, environmental factors, and other factors. Traffic safety, operating speed, manoeu vrability, driver comfort, and service volume are all impacted by pavement performance. Thus, regular pavement evaluation is a very important part of any pavement management system (PMS). The objective of this research is to evaluate the factors affecting pavement performance. International roughness index (IRI) was considered as an indication of pavement performance. The Long Term Pavement Performance (LTPP) database for the flexible pavement experiments was utilized as the database source. Association analysis was conducted between IRI and different variables. These variables are: age, asphalt thickness, climate type, base type, base thickness, and subgrade type. Chi-square test was performed and the result was that all independent variables were significant. Also, odds ratio was determined for different factors and the results was that all variable were significant except subgrade type, some of climate types, and 16 inch base thickness. The latest pavement performance model which correlates IRI with different variables was studied and validated. The independent variables for this model were initial IRI, age, fatigue cracks, transverse cracks, and rutting. This model has R2 = 0.57. This research validated this model using extended LTPP data. The result of this validation was that R2 became 0.46. So, this model calibrated by the same independent variables and the result was that R2 became 0.50. The new models were developed by adding longitudinal cracks to independent variables using linear
regression and machine learning techniques. Multiple linear regression was used to develop a general model for all data and specific models for the different categories of asphalt thickness (4 in and 7 in), climate types (Wet non freeze (WNF), wet freeze (WF), dry non freeze (DNF), and dry freeze (DF)), and subgrade types (fine and coarse). The general model yielded
R2 of 0.60. The specific models were more accurate due to small size of sample that used to develop these models. Finally, Machine learning was used to develop a pavement performance
model. Linear Regression, SGD Regressor, KNN without tuning, KNN with tuning, Random Forest, Xgboost without tuning, and XGboost with tuning were used. The model developed using XGBoost with tuning is the most accurate.